I went to the Ethereum meetup in London, and didn't get to ask this question as I didn't think of it in time.
In the presentation, the example of earthquake insurance was given - that a smart contract could be used to "bet against" or "bet for" an earthquake in a particular place by a particular time, with the money essentially used as insurance is today.
It got me thinking - what if there wasn't an earthquake but enough people thought there had been for some reason?
The earthquake isn't the best example, but humans tend to be herd followers. In 2001, quite a few people thought the NASDAQ stock index should be above 5000. (Plenty of people today think the DOW should be above 16,000.) Enough people given enough credence to what Kim Kardashian does that she is unfortunately a celebrity. There are countless examples of things being 'true' that either shouldn't be or that are debatable.
How will smart contracts handle such things? What if there isn't an earthquake but enough people are convinced (for some crazy reason) that there was and the contract pays out, to have it later disproven?
Is this a function of the contract to correct? That there must be "withdrawal" language? What if I get my ether, spend it all, and then the contract tries to claw it back to find I have none?
How would something like this be handled with the current model (or a future one) of Ethereum?